Comparative analysis of Erk phosphorylation suggests a mixed strategy for measuring phospho-form distributions. Four methods were compared for quantifying the proportion of Erk2 ‘phospho-forms' in differentially phosphorylated samples, revealing excellent agreement between mass spectrometry and nuclear magnetic resonance, but significant discrepancies with western blots.
We compared four methods for quantifying the proportions of phospho-forms, or ‘phospho-form distribution', of a multiply phosphorylated protein, using as an example the MAP kinase Erk2, with two principal phosphorylation sites.Measurements by mass spectrometry (MS) and by nuclear magnetic resonance (NMR) agreed to within 10%, but phospho-specific antibody measurements exhibited semi-quantitative discrepancies with these, sometimes suggesting reverse trends to those found by the biophysical methods.NMR revealed under our conditions that Erk was phosphorylated on four, not two, sites.A combination of peptide-based MS and protein-based MS provided an optimum strategy for determining the 16=24 member phospho-form distribution.
Protein post-translational modification is one of the most significant regulatory mechanisms in cellular physiology, and protein phosphorylation is the most widely studied of these. An individual molecule may be phosphorylatable on multiple residues, allowing it to exist in a multiplicity of combinatorial patterns of modification; with n-sites, there may be 2n such ‘phospho-forms'. Recent work on many distinct types of proteins—ion channels, signalling enzymes, transcription factors, co-activators, circadian clock components—has shown that different phospho-forms may have different downstream effects. The impact of multisite phosphorylation, therefore, is determined by the proportions of the various phospho-forms that are present in the molecular population of the given protein. This ‘phospho-form distribution' is dynamically and collectively regulated by the opposing actions of the relevant kinases and phosphatases. This presents a more challenging perspective than is depicted in the typical cartoon diagram, in which is shown only a single phospho-form, usually the maximally phosphorylated one, and the underlying dynamics of modification and demodification is left implicit.
The present paper sets out to bring this perspective of phospho-forms and phospho-form distributions to a wider biological audience, to compare current methods for measuring them and to discuss the challenges in developing a general strategy applicable to the kinds of proteins typically found in cellular physiology. We chose to analyse the 42 kDa mitogen-activated protein (MAP) kinase Erk2 (Erk). Erk is a paradigmatic signalling protein that is phosphorylated on T and Y residues in a TEY motif within its kinase-activation loop. Because these sites are so close together, multiple methods may be used to detect phospho-forms and we compared four: western blots with phospho-specific antibodies; peptide-based liquid chromatography (LC)/mass spectrometry (MS), in which proteins are first digested into peptide (pepMS); protein-based LC/MS with intact proteins (proMS); and nuclear magnetic resonance spectroscopy (NMR). To provide a stringent comparison, we used specific kinases and phosphatases to prepare four samples of Erk in distinct states of phosphorylation and sought to determine the proportion of each of the four phospho-forms in each of the four samples.
We found excellent agreement, to within 10%, between the various biophysical methods, pepMS, proMS and NMR, despite the experiments being carried out in three different laboratories on two different continents. NMR also revealed the presence of two additional phosphorylations on one of the samples, which we identified by MS as serine phosphorylations. We determined most of the corresponding 16 member phospho-form distribution using a combination of pepMS and proMS. To our surprise, however, we found significant semi-quantitative discrepancies between the biophysical and the immunological methods, despite using the LICOR method of ratiometric fluorescent imaging for western blotting. For instance, a phospho-specific antibody may indicate that sample one has a higher proportion of a certain phospho-form than sample two, but pepMS measurements may sometimes indicate the reverse (compare Figures 1D and 2C). We found similar discrepancies with an alternative set of samples, prepared differently (compare Figure 3A and B), and after spiking western blots with whole-cell lysate (Figure 3C) and after using chemiluminescence and CCD imaging in place of fluorescent detection.
Antibodies are usually characterised for the purposes of quantitative measurement by titration against the same sample, but molecular recognition between antibody and antigen is an emergent property of the biological context. In the comparisons made here, the same phospho-form is being examined in different samples and hence in the context of different phospho-form distributions. The antibody sees not only the phospho-epitope that is its nominal target but also differing amounts of other phospho-epitopes against which it may have a range of cross-reactivities. Such a context-dependent interaction may be one reason for the surprising discrepancies that were found. If so, it exemplifies one of our central themes: it is not any single phospho-form that determines the downstream response, in this case of an antibody; it is the entire phospho-form distribution. While antibodies remain unrivalled for protein detection sensitivity amidst complex cellular backgrounds, our results suggest that care is required in using them for quantitative comparisons of post-translational modifications.
If sites can be localised to a single peptide, pepMS with both LC and MS offers good opportunities for separating distinct phospho-forms, including isobaric ones having the same molecular weight. However, such measurements are not quantitative because distinct phospho-forms may ionise and ‘fly' with different efficiencies; isotopically-labelled phospho-peptide internal standards are required for accurate measurements. Moreover, sites on distinct peptides can no longer be correlated with each other. Measurements with intact protein, as in proMS, avoid both problems but are also less sensitive to LC separation, allowing only the distribution of isobaric forms to be determined. The use of multiple samples, prepared with specific phosphatases, can also be informative. The combination of pepMS and proMS offers a hybrid strategy that represents a good balance between coverage and resolution and holds out promise as a general approach for proteins with small numbers of sites. NMR has certain inherent limitations, being unable to detect correlations between phosphorylations that are not close together and being better able to detect phosphorylation on serine and threonine than on tyrosine. However, it has the advantage, as does proMS, of not being biased by prior expectations as to where modifications are expected, illustrated by its uncovering of the two additional phosphorylations on Erk. While NMR's specialised requirements make it less feasible as a general methodology, it holds out the promise of real-time measurements both in vitro and in intact cells.
The problem of quantifying phospho-form distribution remains very challenging as the number of phosphorylated sites increases but quantitative information is now becoming available for proteins such as Erk that are commonly encountered in cellular physiology.
The functional impact of multisite protein phosphorylation can depend on both the numbers and the positions of phosphorylated sites—the global pattern of phosphorylation or ‘phospho-form'—giving biological systems profound capabilities for dynamic information processing. A central problem in quantitative systems biology, therefore, is to measure the ‘phospho-form distribution': the relative amount of each of the 2n phospho-forms of a protein with n-phosphorylation sites. We compared four potential methods—western blots with phospho-specific antibodies, peptide-based liquid chromatography (LC) and mass spectrometry (MS; pepMS), protein-based LC/MS (proMS) and nuclear magnetic resonance spectroscopy (NMR)—on differentially phosphorylated samples of the well-studied mitogen-activated protein kinase Erk2, with two phosphorylation sites. The MS methods were quantitatively consistent with each other and with NMR to within 10%, but western blots, while highly sensitive, showed significant discrepancies with MS. NMR also uncovered two additional phosphorylations, for which a combination of pepMS and proMS yielded an estimate of the 16-member phospho-form distribution. This combined MS strategy provides an optimal mixture of accuracy and coverage for quantifying distributions, but positional isomers remain a challenging problem.